Disposition Sequence

1library(yamaa)23yamaa_spec("spec.yaml")$run()$output

1import yamaa23yamaa.spec("spec.yaml").run().output

1# Reference solution for the yamaa benchmark sdtm-ds-sequence (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)9library(stringr)1011ds_raw <- read_csv(12  "/app/input/ds_raw.csv",13  col_types = cols(.default = col_character())14)1516complete_date <- function(x) {17  x <- str_trim(x)18  if (is.na(x) || x == "") {19    return(as.Date(NA))20  }21  parts <- str_split(x, "-", simplify = TRUE)22  n <- ncol(parts)23  if (n == 3 && nchar(parts[1]) == 4) {24    return(suppressWarnings(as.Date(x)))25  }26  if (n == 2 && nchar(parts[1]) == 4) {27    y <- suppressWarnings(as.integer(parts[1]))28    m <- suppressWarnings(as.integer(parts[2]))29    if (is.na(y) || is.na(m) || m < 1 || m > 12) {30      return(as.Date(NA))31    }32    return(suppressWarnings(as.Date(sprintf("%04d-%02d-15", y, m))))33  }34  if (n == 1 && nchar(x) == 4) {35    y <- suppressWarnings(as.integer(x))36    if (is.na(y)) {37      return(as.Date(NA))38    }39    return(suppressWarnings(as.Date(sprintf("%04d-06-15", y))))40  }41  as.Date(NA)42}4344ds <- ds_raw |>45  rowwise() |>46  mutate(47    DOMAIN = "DS",48    STUDYID = STUDY,49    USUBJID = PATNUM,50    DSCAT = ifelse(DSDECOD == "RANDOMIZED", "PROTOCOL MILESTONE", "DISPOSITION EVENT"),51    DSDTC = complete_date(DSDTCOL)52  ) |>53  ungroup() |>54  arrange(USUBJID, DSDTC, DSDECOD) |>55  group_by(USUBJID) |>56  mutate(DSSEQ = row_number()) |>57  ungroup() |>58  select(DOMAIN, STUDYID, USUBJID, DSSEQ, DSDECOD, DSCAT, DSDTC)5960dir.create("/app/output", showWarnings = FALSE)61write_csv(ds, "/app/output/ds.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-ds-sequence (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67"""Build the SDTM DS dataset from the collected DS_RAW dataset.89One output record per collected disposition record. Partial collection10dates are completed before use: a year and month becomes the 15th of11that month, and a year alone becomes June 15th of that year. DSSEQ12numbers each subject's records from 1 in order of the completed13collection date, earliest first; records with no date come after the14subject's dated records.15"""1617import csv18import re19from datetime import date2021INPUT_PATH = "/app/input/ds_raw.csv"22OUTPUT_PATH = "/app/output/ds.csv"2324FULL_DATE = re.compile(r"^(\d{4})-(\d{2})-(\d{2})$")25YEAR_MONTH = re.compile(r"^(\d{4})-(\d{2})$")26YEAR_ONLY = re.compile(r"^(\d{4})$")272829def complete_date(raw):30    """Return the completed ISO date string, or None if unreadable."""31    if raw is None:32        return None33    text = raw.strip()34    if not text:35        return None36    match = FULL_DATE.match(text)37    if match:38        year, month, day = (int(part) for part in match.groups())39    else:40        match = YEAR_MONTH.match(text)41        if match:42            year, month = (int(part) for part in match.groups())43            day = 1544        else:45            match = YEAR_ONLY.match(text)46            if match:47                year = int(match.group(1))48                month, day = 6, 1549            else:50                return None51    try:52        completed = date(year, month, day)53    except ValueError:54        return None55    return completed.isoformat()565758def main():59    with open(INPUT_PATH, newline="", encoding="utf-8") as handle:60        rows = list(csv.DictReader(handle))6162    records = []63    for position, row in enumerate(rows):64        outcome = (row.get("DSDECOD") or "").strip()65        dsdtc = complete_date(row.get("DSDTCOL"))66        records.append(67            {68                "DOMAIN": "DS",69                "STUDYID": row.get("STUDY") or "",70                "USUBJID": row.get("PATNUM") or "",71                "DSDECOD": outcome,72                "DSCAT": (73                    "PROTOCOL MILESTONE"74                    if outcome == "RANDOMIZED"75                    else "DISPOSITION EVENT"76                ),77                "DSDTC": dsdtc or "",78                "_sort_date": dsdtc,79                "_position": position,80            }81        )8283    subjects = []84    for record in records:85        if record["USUBJID"] not in subjects:86            subjects.append(record["USUBJID"])8788    output = []89    for subject in subjects:90        own = [r for r in records if r["USUBJID"] == subject]91        own.sort(92            key=lambda r: (93                r["_sort_date"] is None,94                r["_sort_date"] or "",95                r["_position"],96            )97        )98        for index, record in enumerate(own, start=1):99            record["DSSEQ"] = index100            output.append(record)101102    fieldnames = [103        "DOMAIN",104        "STUDYID",105        "USUBJID",106        "DSSEQ",107        "DSDECOD",108        "DSCAT",109        "DSDTC",110    ]111    with open(OUTPUT_PATH, "w", newline="", encoding="utf-8") as handle:112        writer = csv.DictWriter(handle, fieldnames=fieldnames)113        writer.writeheader()114        for record in output:115            writer.writerow({name: record[name] for name in fieldnames})116117118if __name__ == "__main__":119    main()

1/* opensas reference solution for the yamaa benchmark sdtm-ds-sequence.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data ds_raw;6  length STUDY PATNUM INSTANCE DSDECOD DSDTCOL $1024;7  infile "/app/input/ds_raw.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDY : $char1024. PATNUM : $char1024. INSTANCE : $char1024. DSDECOD : $char1024. DSDTCOL : $char1024.;9run;1011data prepared; set ds_raw; length DOMAIN $2 STUDYID USUBJID DSCAT DSDTC $1024; DOMAIN='DS'; D=.; DSDTC=''; STUDYID=STUDY; USUBJID=PATNUM;12if DSDECOD='RANDOMIZED' then DSCAT='PROTOCOL MILESTONE'; else DSCAT='DISPOSITION EVENT';13if lengthn(DSDTCOL)=10 then D=input(DSDTCOL,?? yymmdd10.); else if lengthn(DSDTCOL)=7 then D=input(cats(DSDTCOL,'-15'),?? yymmdd10.); else if lengthn(DSDTCOL)=4 then D=input(cats(DSDTCOL,'-06-15'),?? yymmdd10.);14if not missing(D) then DSDTC=put(D,yymmdd10.); NODATE=missing(D); run;15proc sort data=prepared; by STUDYID USUBJID NODATE D; run;16data result; set prepared; by STUDYID USUBJID; retain DSSEQ;17  if first.USUBJID then DSSEQ=0;18  DSSEQ+1;19run;2021proc sql;22  create table final as select DOMAIN, STUDYID, USUBJID, DSSEQ, DSDECOD, DSCAT, DSDTC from result;23quit;24proc export data=final outfile="/app/output/ds.csv" dbms=csv replace; run;